Prognostic biomarkers in oral squamous cell carcinoma: a systematic review

Over the years, several tumor biomarkers have been suggested to foresee the prognosis of oral squamous cell carcinoma (OSCC) patients. Here, we present a systematic review to identify, evaluate and summarize the evidence for OSCC reported markers. Eligible studies were identified through a literature search of MEDLINE/PubMed until January 2016. We included primary articles reporting overall survival, disease-free survival and cause-specific survival as outcomes. Our findings were analysed using REporting recommendations for tumor MARKer prognostic studies (REMARK), QuickGo tool and SciCurve trends. We found 41 biomarkers, mostly proteins evaluated by immunohistochemistry. The selected studies are of good quality, although, any study referred to a sample size determination. Considering the lack of follow-up studies, the molecules are still potential biomarkers. Further research is required to validate these biomarkers in well-designed clinical cohort-based studies.


INTRODUCTION
Oral squamous cell carcinoma (OSCC) is the most common malignancy of the head and neck (excluding nonmelanoma skin cancer), with more than 300,000 new cases reported annually worldwide [1]. The disease has a high morbidity rate (37.8%) five years after diagnosis (http://www.cancer.gov/statistics/find -2003-2009 data); despite the progress in research and therapy, survival has not improved significantly in the last few decades [2]. The search for prognostic markers represents a continuing challenge for biomedical science.
A cancer biomarker may be a molecule secreted by a tumor cell or a specific response of the body to the presence of cancer [3]. Biomarkers can be used for patient assessment in multiple clinical settings, including estimating the risk of disease and distinguishing benign from malignant tissues [4]. Cancer biomarkers can be classified based on the disease state, including predictive, diagnosis and prognosis biomarkers [5]. A prognostic biomarker informs about a likely cancer outcome (e.g., overall survival, disease-free survival, and cause-specific survival) independent of treatment received [6]. From the identification of a promising biomarker to its clinical use, there is a long pathway involving many complicated hurdles, such as estimating the number of patients needed for the validation phase and statistical validation, among others [7,8]. This validation and qualification are responsible for linking the promising biomarker with a biological process to clinical endpoints [9].
Considering several tumor biomarkers have been suggested to predict the prognosis of OSCC patients, we performed a systematic review, which is widely accepted as a "gold standard" in medicine based on evidence [10], to identify, evaluate and summarize the evidence for OSCC reported markers.

MATERIALS AND METHODS
We performed a systematic review to conduct this investigation. The independent variables were prognostic biomarkers; the dependent variables were OSCC outcomes.

Search strategy
A systematic review allows critical analysis of multiple research studies. Aiming to answer the question "what are the biomarkers of OSCC?", a systematic literature search based on keywords was performed. As PubMed comprises more than 26

Inclusion criteria
Articles were included based on a previously published protocol [11]. Briefly, studies were selected if they examined the impact of a potential biological marker on at least one of the features in OSCC patients: OS, DFS or CSS. These definitions were assessed among the selected papers.
In addition, if a study was focused on isolated or combined (multiple) tumor biomarkers, it must have been subjected to multivariable analysis with one or more additional variables.

Exclusion criteria
Articles were excluded from the present review for the following reasons: i) lack of the terms "oral cancer" and "risk" in their titles, abstracts or keywords; ii) absence of risk ratios and iii) unclear defining criteria for groups and variables.

Potential prognostic biomarker
To determine whether a biomarker is potentially prognostic, the selected articles showed: i) a formal test (binary logistic regression or Cox proportional hazards model) and ii) a statistically significantly association between the biomarker and outcome [6]. The computed risk (odds ratio, OR or hazard ratio, HR) was reported as the risk of a specific outcome from the biomarker group versus the reference group, with OR/HR>1 indicating increased risk and OR/HR<1 indicating decreased risk.

Data extraction
One investigator reviewed all the eligible studies and carefully extracted the study characteristics, including the article citation information, biomarker name and classification, condition or outcome, laboratory technique, sample size, number of clinical outcomes, status of biomarker expression, statistical test method, computed risk and its p-value and 95% confidence interval (CI). The main biological processes in which the biomarkers are involved were obtained using QuickGo (http://www.ebi.ac.uk/QuickGO).

Quality assessment
Quality assessment was performed in duplicate for each eligible study by three independent reviewers using operationalized prognostic biomarker reporting the REMARK guidelines [12] and extracted details on 20 items. The inter-observer agreement was evaluated using Kappa statistics.

Publication trends
To observe the publication trends in the selected potential OSCC biomarkers, we searched the scholarly literature in SciCurve Open (http://www.scicurve.com). SciCurve Open is a search engine that transforms a systematic literature review into an interactive and comprehensible environment [13].

Studies searching for OSCC biomarkers: proteins are the most analysed molecules
The keyword search strategy identified 403 suitable abstracts, from which 320 were excluded by reviewing the title and abstract during the screen because they did not meet the eligibility criteria. Full text articles were obtained for 83 studies (34 with single markers and 49 with multiple or combined markers).
Forty-five of these articles were excluded for different reasons, including: out of goal (3 articles), unavailability online (2 articles), lack of multivariable analysis (18 articles) and model inconsistencies (22 articles). Figure 1 shows a PRISMA diagram for this review (for details, see Supplemental file S1).
The selected studies were screened, and specific study characteristics and remarks were recorded. These parameters are summarized in Table 1 (the article context is grouped according to the hallmarks of cancer [14]). Thirty-eight papers examined 41 biomarkers . Most of them were proteins determined using immunohistochemistry (IHC) in paraffin-embedded tissues ( Figure 2). For complete details, see Supplemental file S1.

Quality of study reports: studies do not clear determine the sample size
The result of this agreement was 0.87, which is classified as almost perfect. Differences were resolved by consensus. Most study analyses reported details of the objective/hypothesis, patient source, population characteristics, assay method, cut-off point, and relationship of the potential marker to standard prognostic variables, as well as discussed the implications for future research and clinical value (for details, see Supplemental file S2). Notably, no study referred to a statistical sample size, which is key for biomarker validation.

Proposed OSCC biomarkers
None of the studied molecules presented an analysis of validation, so we called them "potential biomarkers". A narrative review of the proposed biomarkers is presented in Table 3.

Trends: potential biomarkers with more publications and citations
To explore the publication trends in our OSCC potential protein biomarkers, we searched the scholarly literature in SciCurve Open. SciCurve uses PubMed's library of 23 million references to generate visually pleasing graphs and curves that help grasp trends in the literature [53]. It is associated with the following main functionalities: publications, citations, most prolific authors and countries.
According to Figure 3, MMP-2 is the most researched field, followed by MMP-1, cadherin-1 and mucin-1. The countries with the largest contributions are the USA, Japan and China.

DISCUSSION
We have summarized the results on the association between biomarkers and oral cancer outcomes using a systematic review. Overall, our results suggest 41 prognostic molecules involved with OSCC endpoints. These markers may be candidates for long-term studies.
OSCC is the most relevant epithelial malignancy for dental surgeons. It has late clinical detection and poor prognosis, and the available therapeutic alternatives are highly expensive and disfiguring [54].
OSCC is a very complex subtype of cancer with high heterogeneity [55]. Several risk factors are implicated in its aetiology, among which tobacco, alcohol, viruses and diet are highlighted [2]. These factors related to genetic inheritance may have a carcinogenic effect on the normal cells of the respiratory and digestive systems. This type of carcinoma can occur anywhere in the mouth, although the most affected sites are the tongue, lower lip and mouth floor [2,56].
These regions are great facilitators of carcinoma spreading to regional lymph nodes and/or distant organs [57]. At present, the diagnosis of OSCC is based on comprehensive clinical examination and histological analysis of suspicious areas [58]. Recently, The Cancer Genome Atlas (TCGA) showed that a large dataset of proteomics/genomics did not improve the prognosis potential of classic clinical variables in patients with different types of cancer [59]. Some studies seeking biomarkers in oral cancer are still in the discovery phase, requiring validation to be accepted in clinical practice.
Currently, biomarkers are a subject of particular interest because they may represent the most important part in the diagnosis step. In the future, specific and personalised diagnostics can guide treatment against the disease and consequently improve the chance of curing the disease.
In response to the need for tumor biomarkers for OSCC that can be readily evaluated in The malignant progression to OSCC is characterized by the acquisition of progressive and uncontrolled growth of tumor cells. Predicting whether premalignant lesions will progress to cancer is crucial to make appropriate treatment decisions. The first detectable clinical changes that can indicate that an epithelium is on the way to establish OSCC is the occurrence of malignant disorders, including leukoplakia (most common) [2]. In this context, we emphasize the results associated with Rho GTPase-activating protein 7, retinal dehydrogenase 1/prominin-1 (combined biomarkers), podoplanin, cortactin/focal adhesion kinase 1 (combined biomarkers) and catenin delta-1. These proteins show a potential role as a marker of oral cancer risk and malignant transformation [17, 26-28, 39, 40, 42].
There are thousands of papers reporting cancer biomarker discovery, but only few clinically useful biomarkers have been successfully validated for routine clinical practice [62]. Quality assessment tools have been developed for prognostic studies to help identify study biases and causes of heterogeneity when performing meta-analysis. We chose to use the REMARK reporting guidelines, which provide a useful start for assessing tumor prognostic biomarkers (all included studies were prognostic). We found that the investigations reported an average of 19 of 20 REMARK items. However, all studies failed to report the sample size calculation. In the absence of this calculation, the findings of each research should be interpreted with caution [63]. The sample size requirements that allow the identification of a benefit beyond existing biomarkers are even more demanding [64].
In our review, none of the articles that created prediction models had internal or external validation. In general, studies recruited cases of OSCC from a clinical setting as well as controls without a clearly defined diagnosis. Under this circumstance, any differences in the biomarker levels between OSCC patients and controls could simply reflect individual differences rather than cancer-related differences. The lack of biomarker validation strategies and standard operating procedures for sample selection in the included studies represent an important pitfalls and limitations, leading us to use the term "potential biomarkers" instead of biomarker in our article title.
It is important to highlight that our research searched only one database, which means that only studies available in MEDLINE were included. Additionally, due to the heterogeneity among the studies, a meta-analysis that combined the results of different studies could not be performed.
In addition, our research included results from observational studies, and their evaluation may have been problematic if the confounder variables were not adjusted because they were not measured [65].

CONCLUSION
Recent research in OSCC has identified a multitude of potential markers that have a significant role in prognosis. In this systematic review, despite the inherent limitations, we identified several potential biomarkers of particular interest that appear to carry prognostic significance. Considering the validation step as a process of assessing the biomarker and its measurement performance characteristics, and determine the range of conditions under which this biomarker can provide reproducible data [9], our results show biomarkers in the discovery phase, thereby leading us to call them OSCC "potential biomarkers". Nevertheless, it is urgent to apply validation methods to provide clinically useful oral cancer biomarkers.    Immunohistochemical detection appears to improve diagnosis and to provide prognostic information in addition to the TNMsystem and histological grade of OSCC. Fillies T et al., 2005   HIF1A (-) Germany.
Overexpression is an indicator of favorable prognosis in T1 and T2 SCC of the oral floor. Node negative patients lacking expression may therefore be considered for adjuvant radiotherapy. High levels of both preoperative SERPINB3 and CRP levels act as a predictor for DFS and OS.

Tumor-promoting inflammation
The articles are grouped according to the hallmarks of cancer. *UniProt Knowledgebase or common name. HGNC name between parentheses. (+) Up-regulated/overexpressed, (-) Down-regulated/down-expressed, CSS, cause-specific survival; OS, overall survival; DFS, disease free survival; IHC, immunohistochemistry; FISH, fluorescence in situ hybridization.  Cell cycle, cell proliferation. Marker of the growth fraction for a certain cell population [1]. The labelling index is considered one of the best prognostic factors of the survival rate and recurrence [2]. CDKN2A Cell cycle, cell cycle arrest. This gene is frequently mutated or deleted in a wide variety of tumors, and is known to be an important tumor suppressor gene [3]. HPV16 High-risk HPV type. Is emerging as an important factor in the rise of oropharyngeal tumors affecting non-smokers in developed countries. Patients with HPV(+) tumors demonstrated favorable outcomes compared to TP53 mutants and 11q13/CCND1-amplified tumors [4]. DLC1 Negative regulation of cell proliferation and migration.
Acts as a tumour suppressor in a number of common cancers, including liver cancer [5].

CYR61
Regulation of cell growth and adhesion.
Can function as an oncogene or a tumour suppressor, depending on the origin of the cancer [6].

TP53
Cell cycle, cell cycle arrest. Tumor-suppressor protein. Mutations in this gene are associated with a variety of human cancers [3]. CA9 Response to hypoxia. Is the most widely expressed gene in response to hypoxia. Its role in intracellular pH maintenance represents the means by which cancer cells adapt to the toxic conditions of the extracellular environtment [7] CCND1 Cell cycle, cell division. Is frequently deregulated in cancer and is a biomarker of cancer phenotype and disease progression [8]. EGFR Positive regulation of cell proliferation.
EGFR overexpression is a significant finding in cancer, particularly in head and neck cancer, where it is also associated with a poor prognosis [9].

RB1
Cell cycle, cell cycle arrest. Tumor-suppressor protein. Defects in this gene are a cause of childhood cancer retinoblastoma (RB), bladder cancer, and osteogenic sarcoma [3].

MYC
Positive regulation of cell proliferation.
Its oncogenic reputation stems from its frequent deregulation in a host of human cancers and from a suite of activities that place this protein at the nexus of cell growth, proliferation, metabolism, and genome stability [10]. ALDH1A1 Ethanol oxidation. Play a key role in the regulation of growth and differentiation of both normal tissue stem cells and cancer stem cells [11]. PROM1 Retina layer formation. Maintaining stem cell properties by suppressing differentiation [3]. S100-A2 Endothelial cell migration.
In epithelial tissue, S100-A2 expression is decreased remarkably in tumours compared with normal specimens [12]. S100-A2 promotes p53 transcriptional activity, and its loss of expression has been associated with a poorer prognosis and shorter survival [13]. CDC20 Cell cycle, positive regulation of cell proliferation.
The role of CDC20 expression in tumours is not known, but many studies have reported that CDC20 regulates apoptosis, leading to genetically instability [14] MAP1LC3A Autophagy. Strong positive expression in the peripheral area of pancreatic cancer tissue had a shorter overall and disease-free survival; correlations with tumour size, poor differentiation, blood vessel infiltration and tumour necrosis were noted [15]. FAS Apoptotic process. Cancer cells can never lose FAS or FASLG. FAS and/or FASLG expression promotes tumor growth and favors the establishment of tumor metastases [16]. FASLG HMOX1 Angiogenesis. Many human tumours produce HMOX1, and its expression is usually higher in cancer cells than in surrounding healthy tissues [17]. PDPN Lymphangiogenesis. Is commonly used in the identification of lymphatic endothelial differentiation in vascular endothelial neoplasms and lymphatic invasion by tumours [18]. Recent evidence have identified podoplanin as a marker of cancer-associated fibroblasts [19]. CTTN Cell motility and focal adhesion assembly.
Is overexpressed in breast cancer and squamous cell carcinomas of the head and neck [3].

Promotes tumor progression and metastasis through effects on cancer cells, as well as stromal cells of the tumor microenvironment [20] MUC4
Cell adhesion. An aberrant expression of MUC4 has been reported in various carcinomas [21]. CTNND1 Cell adhesion. Evidence is emerging that complete loss, downregulation or mislocalization of CTNND1 correlates with the progression of different types of human tumours [22]. ACTA2 Mesenchyme migration. Patients with lung adenocarcinomas and high ACTA2 expression showed significantly enhanced distant metastasis and unfavorable prognosis [23].

MMP1
Proteolysis. Imbalance between matrix metalloproteinases and their inhibitors play the important role in progression of head and neck cancer [24].

MMP2
Angiogenesis, response to hypoxia and proteolysis. VIM Movement of cell or subcellular component.
Has been recognized as a marker for epithelial-mesenchymal transition. Overexpression in cancer correlates well with accelerated tumor growth, invasion, and poor prognosis [25]. CDH1 Cell adhesion. Loss of function of this gene is thought to contribute to cancer progression by increasing proliferation, invasion, and/or metastasis [3]. VCAN Cell adhesion. Is strongly associated with a poor outcome for many different cancers. Depending on the cancer nature, is expressed either by cancer cells themselves or by stromal cells surrounding the tumour [26]. AMFR Movement of cell or subcellular component.
Is a tumor motility-stimulating protein secreted by tumor cells [3].

MUC1
DNA damage response, signal transduction by p53 class mediator resulting in cell cycle arrest Is aberrantly glycosylated and overexpressed in various epithelial cancers and plays a crucial role in the progression of the disease [27]. MUC1 is often used as a diagnostic marker for metastatic progression [28].
Up-regulates the expression of proteins that promote angiogenesis, anaerobic metabolism, and many other survival pathways [29]. SLC2A1 Glucose transport. Was significantly correlated with depth of invasion and clinical stage in patients with gastric cancer [30].

IL4R
Immune system process and regulation of cell proliferation The IL4/IL4R signaling axis is a strong promoter of pro-metastatic phenotypes in epithelial cancer cells including enhanced migration, invasion, survival, and proliferation [31]. IL13RA1 Cell surface receptor signaling pathway.
Glioblastoma samples presented higher IL13RA1 and IL13RA2 expression levels compared to lower grades astrocytomas and non-neoplastic cases [32]. CXCL8 Angiogenesis, movement of cell or subcellular component and chemotaxis Neovascularisation is now recognised as a critical function of CXCL8 in the tumour microenvironment [33].

CD163
Inflammatory response. Could be used as a general anti-inflammatory myeloid marker with prognostic impact for breast cancer patients [34]. MPO Defense response. Myeloperoxidase-positive cell infiltration in colorectal carcinogenesis is an indicator of colorectal cancer risk [35]. SERPINB3 Positive regulation of cell proliferation.
Promotes oncogenesis and epithelial-mesenchymal transition [36] CRP Inflammatory response. Patients with a high baseline CRP had a greater risk of early death compared with those with low CRP levels [37]. *HGNC database recommended names were used. **Representative processes from QuickGo (http://www.ebi.ac.uk/QuickGO).